Source code for glue.core.roi

import copy

import numpy as np
from matplotlib.patches import Ellipse, Polygon, Rectangle, Path as MplPath, PathPatch
from matplotlib.transforms import IdentityTransform, blended_transform_factory

from glue.core.component import CategoricalComponent
from glue.core.exceptions import UndefinedROI
from glue.utils import points_inside_poly, iterate_chunks, rotation_matrix_2d


np.seterr(all='ignore')


__all__ = ['Roi', 'RectangularROI', 'CircularROI', 'PolygonalROI',
           'AbstractMplRoi', 'MplRectangularROI', 'MplCircularROI',
           'MplPolygonalROI', 'MplXRangeROI', 'MplYRangeROI',
           'XRangeROI', 'RangeROI', 'YRangeROI', 'VertexROIBase',
           'CategoricalROI', 'EllipticalROI']

PATCH_COLOR = '#FFFF00'
SCRUBBING_KEY = 'control'


def aspect_ratio(axes):
    """Returns the pixel height / width of a box that spans 1 data unit in `x` and `y`"""
    width = axes.get_position().width * axes.figure.get_figwidth()
    height = axes.get_position().height * axes.figure.get_figheight()
    xmin, xmax = axes.get_xlim()
    ymin, ymax = axes.get_ylim()
    return height / width / (ymax - ymin) * (xmax - xmin)


def data_to_norm(axes, x, y):
    xy = np.column_stack((np.asarray(x).ravel(), np.asarray(y).ravel()))
    pixel = axes.transData.transform(xy)
    norm = axes.transAxes.inverted().transform(pixel)
    return norm


def data_to_pixel(axes, x, y):
    xy = np.column_stack((np.asarray(x).ravel(), np.asarray(y).ravel()))
    return axes.transData.transform(xy)


def pixel_to_data(axes, x, y):
    xy = np.column_stack((np.asarray(x).ravel(), np.asarray(y).ravel()))
    return axes.transData.inverted().transform(xy)


def pixel_to_axes(axes, x, y):
    xy = np.column_stack((np.asarray(x).ravel(), np.asarray(y).ravel()))
    return axes.transAxes.inverted().transform(xy)


[docs]class Roi(object): # pragma: no cover """ A geometrical 2D region of interest. Glue uses ROIs to represent user-drawn regions on plots. There are many specific sub-classes of Roi, but they all have a ``contains`` method to test whether a collection of 2D points lies inside the region. ROI bounds are generally designed as being exclusive (that is, points situated exactly on a border are considered to lie outside the region). """
[docs] def contains(self, x, y): """ Test which of a set of (`x`, `y`) points fall within the region of interest. Parameters ---------- x : float or array-like `x` coordinate(s) of point(s). y : float or array-like `y` coordinate(s) of point(s). Returns ------- inside : bool or `~numpy.ndarray` An boolean iterable, where each element is `True` if the corresponding (`x`, `y`) tuple is inside the Roi. Raises ------ UndefinedROI If not defined. """ raise NotImplementedError()
[docs] def contains3d(self, x, y, z): """ Test which of a set of projected (`x`, `y`, `z`) points fall within the linked 2D region of interest. Parameters ---------- x : :class:`~numpy.ndarray` Array of `x` locations y : :class:`~numpy.ndarray` Array of `y` locations z : :class:`~numpy.ndarray` Array of `z` locations Returns ------- :class:`~numpy.ndarray` A boolean array, where each element is `True` if the corresponding (`x`, `y`, `z`) tuple is projected inside the associated 2D Roi. Raises ------ UndefinedROI If not defined. """ raise NotImplementedError()
[docs] def center(self): """Return the (`x`, `y`) coordinates of the ROI center""" raise NotImplementedError()
[docs] def move_to(self, x, y): """Translate the ROI to a center of (`x`, `y`)""" raise NotImplementedError()
[docs] def defined(self): """Returns `True` if the ROI is defined""" raise NotImplementedError()
[docs] def to_polygon(self): """ Returns vertices `vx`, `vy` of a polygon approximating the Roi, where each is an array of vertex coordinates in `x` and `y`. """ raise NotImplementedError()
[docs] def rotate_to(self, theta): """ Rotate anticlockwise around center to position angle theta. Parameters ---------- theta : float Angle of anticlockwise rotation around center in radian. """ raise NotImplementedError()
[docs] def rotate_by(self, dtheta, **kwargs): """ Rotate the Roi around center by angle dtheta. Parameters ---------- dtheta : float Change in anticlockwise rotation angle around center in radian. """ self.rotate_to(getattr(self, 'theta', 0.0) + dtheta, **kwargs)
[docs] def copy(self): """Return a clone of the Roi""" return copy.copy(self)
[docs] def transformed(self, xfunc=None, yfunc=None): """A transformed version of the Roi""" raise NotImplementedError()
class PointROI(Roi): def __init__(self, x=None, y=None): self.x = x self.y = y def contains(self, x, y): return False def move_to(self, x, y): self.x = x self.y = y def defined(self): try: return np.isfinite([self.x, self.y]).all() except TypeError: return False def center(self): return self.x, self.y def reset(self): self.x = self.y = None
[docs]class RectangularROI(Roi): """ A 2D rectangular region of interest. Parameters ---------- xmin, xmax : float, optional `x` coordinates of left and right edge. ymin, ymax : float, optional `y` coordinates of lower and upper edge. theta : float, optional Angle of anticlockwise rotation around center in radian. Notes ----- The input and ``update_limits()`` parameters specify the `x` and `y` edge positions *before* any rotation is applied; the size always remains :math:`width` = `xmax` - `xmin`; `height` = `ymax` - `ymin`. """ def __init__(self, xmin=None, xmax=None, ymin=None, ymax=None, theta=None): super(RectangularROI, self).__init__() self.xmin = xmin self.xmax = xmax self.ymin = ymin self.ymax = ymax self.theta = 0 if theta is None else theta def __str__(self): if self.defined() and self.theta == 0: return f"x=[{self.xmin:.3f}, {self.xmax:.3f}], y=[{self.ymin:.3f}, {self.ymax:.3f}]" elif self.defined(): return (f"center=({self.center()[0]:.3f}, {self.center()[1]:.3f}), " f"size=({self.width():.3f} x {self.height():.3f}), " f"theta={self.theta:.3f} radian") else: return "Undefined Rectangular ROI"
[docs] def center(self): return self.xmin + self.width() / 2, self.ymin + self.height() / 2
[docs] def move_to(self, x, y): cx, cy = self.center() dx = x - cx dy = y - cy self.xmin += dx self.xmax += dx self.ymin += dy self.ymax += dy
[docs] def rotate_to(self, theta): self.theta = 0 if theta is None else theta
[docs] def transpose(self, copy=True): if copy: new = self.copy() new.xmin, new.xmax = self.ymin, self.ymax new.ymin, new.ymax = self.xmin, self.xmax return new self.xmin, self.ymin = self.ymin, self.xmin self.xmax, self.ymax = self.ymax, self.xmax
[docs] def corner(self): return (self.xmin, self.ymin)
[docs] def width(self): return self.xmax - self.xmin
[docs] def height(self): return self.ymax - self.ymin
[docs] def contains(self, x, y): if not self.defined(): raise UndefinedROI if not isinstance(x, np.ndarray): x = np.asarray(x) if not isinstance(y, np.ndarray): y = np.asarray(y) if np.isclose(self.theta % np.pi, 0.0, atol=1e-9): return (x > self.xmin) & (x < self.xmax) & (y > self.ymin) & (y < self.ymax) elif np.isclose(self.theta % (np.pi / 2), 0.0, atol=1e-9): xc, yc = self.center() xext = self.height() / 2 yext = self.width() / 2 return (x > xc - xext) & (x < xc + xext) & (y > yc - yext) & (y < yc + yext) inside = np.zeros_like(x, dtype=bool) bounds = self.to_polygon() xc, yc = self.center() keep = ((x >= bounds[0].min()) & (x <= bounds[0].max()) & (y >= bounds[1].min()) & (y <= bounds[1].max())) x = x[keep] - xc y = y[keep] - yc shape = (2,) + x.shape x, y = (rotation_matrix_2d(-self.theta) @ [x.flatten(), y.flatten()]).reshape(shape) inside[keep] = (abs(x) <= self.width() / 2) & (abs(y) <= self.height() / 2) return inside
[docs] def update_limits(self, xmin, ymin, xmax, ymax): """Update the limits (edge positions) of the rectangle before rotation""" self.xmin = min(xmin, xmax) self.xmax = max(xmin, xmax) self.ymin = min(ymin, ymax) self.ymax = max(ymin, ymax)
[docs] def reset(self): """Reset the rectangular region""" self.xmin = None self.xmax = None self.ymin = None self.ymax = None
[docs] def defined(self): return self.xmin is not None
[docs] def to_polygon(self): """ Returns vertices `vx`, `vy` of the rectangular region represented as a polygon, where each is an array of vertex coordinates in `x` and `y`. """ if self.defined(): if np.isclose(self.theta % np.pi, 0.0, atol=1e-9): return (np.array([self.xmin, self.xmax, self.xmax, self.xmin, self.xmin]), np.array([self.ymin, self.ymin, self.ymax, self.ymax, self.ymin])) else: corners = (np.array([-1, 1, 1, -1, -1]) * self.width() / 2, np.array([-1, -1, 1, 1, -1]) * self.height() / 2) return tuple((rotation_matrix_2d(self.theta) @ corners) + np.array(self.center()).reshape((2, 1))) else: return [], []
[docs] def transformed(self, xfunc=None, yfunc=None): xmin = self.xmin if xfunc is None else xfunc(self.xmin) xmax = self.xmax if xfunc is None else xfunc(self.xmax) ymin = self.ymin if yfunc is None else yfunc(self.ymin) ymax = self.ymax if yfunc is None else yfunc(self.ymax) return RectangularROI(xmin=xmin, xmax=xmax, ymin=ymin, ymax=ymax)
def __gluestate__(self, context): return dict(xmin=context.do(self.xmin), xmax=context.do(self.xmax), ymin=context.do(self.ymin), ymax=context.do(self.ymax), theta=context.do(self.theta)) @classmethod def __setgluestate__(cls, rec, context): return cls(xmin=context.object(rec['xmin']), xmax=context.object(rec['xmax']), ymin=context.object(rec['ymin']), ymax=context.object(rec['ymax']), theta=context.object(rec.get('theta', 0)))
[docs]class RangeROI(Roi): """ A region of interest representing all points within a range in either `x` or `y`. Parameters ---------- orientation : str One of 'x' or 'y', setting the axis on which to apply the range. min : float, optional Start value of the range. max : float, optional End value of the range. """ def __init__(self, orientation, min=None, max=None): super(RangeROI, self).__init__() self.min = min self.max = max self.ori = orientation @property def ori(self): return self._ori
[docs] @ori.setter def ori(self, value): if value in set('xy'): self._ori = value else: raise ValueError("Orientation must be one of 'x', 'y'")
def __str__(self): if self.defined(): return "%0.3f < %s < %0.3f" % (self.min, self.ori, self.max) else: return "Undefined %s" % type(self).__name__
[docs] def range(self): return self.min, self.max
[docs] def center(self): return (self.min + self.max) / 2
[docs] def set_range(self, lo, hi): self.min, self.max = lo, hi
[docs] def move_to(self, center): delta = center - self.center() self.min += delta self.max += delta
[docs] def contains(self, x, y): if not self.defined(): raise UndefinedROI() coord = x if self.ori == 'x' else y return (coord > self.min) & (coord < self.max)
[docs] def transformed(self, xfunc=None, yfunc=None): if self.ori == 'x': vmin = self.min if xfunc is None else xfunc(self.min) vmax = self.max if xfunc is None else xfunc(self.max) else: vmin = self.min if yfunc is None else yfunc(self.min) vmax = self.max if yfunc is None else yfunc(self.max) return RangeROI(orientation=self.ori, min=vmin, max=vmax)
[docs] def reset(self): self.min = None self.max = None
[docs] def defined(self): return self.min is not None and self.max is not None
[docs] def to_polygon(self): if self.defined(): on = [self.min, self.max, self.max, self.min, self.min] off = [-1e100, -1e100, 1e100, 1e100, -1e100] x, y = (on, off) if (self.ori == 'x') else (off, on) return x, y else: return [], []
def __gluestate__(self, context): return dict(ori=self.ori, min=context.do(self.min), max=context.do(self.max)) @classmethod def __setgluestate__(cls, rec, context): if cls is XRangeROI or cls is YRangeROI: return cls(min=rec['min'], max=rec['max']) else: return cls(rec['ori'], min=rec['min'], max=rec['max'])
[docs]class XRangeROI(RangeROI): def __init__(self, min=None, max=None): super(XRangeROI, self).__init__('x', min=min, max=max)
[docs]class YRangeROI(RangeROI): def __init__(self, min=None, max=None): super(YRangeROI, self).__init__('y', min=min, max=max)
[docs]class CircularROI(Roi): """ A 2D circular region of interest. Parameters ---------- xc : float, optional `x` coordinate of center. yc : float, optional `y` coordinate of center. radius : float, optional Radius of the circle. """ def __init__(self, xc=None, yc=None, radius=None): super(CircularROI, self).__init__() self.xc = xc self.yc = yc self.radius = radius
[docs] def contains(self, x, y): if not self.defined(): raise UndefinedROI if not isinstance(x, np.ndarray): x = np.asarray(x) if not isinstance(y, np.ndarray): y = np.asarray(y) return (x - self.xc) ** 2 + (y - self.yc) ** 2 < self.radius ** 2
[docs] def set_center(self, x, y): """Set the center of the circular region""" self.xc = x self.yc = y
[docs] def set_radius(self, radius): """Set the radius of the circular region""" self.radius = radius
[docs] def get_center(self): return self.xc, self.yc
[docs] def get_radius(self): return self.radius
[docs] def reset(self): """Reset the circular region""" self.xc = None self.yc = None self.radius = 0.
[docs] def defined(self): return self.xc is not None and \ self.yc is not None and self.radius is not None
[docs] def to_polygon(self): if not self.defined(): return [], [] theta = np.linspace(0, 2 * np.pi, num=20) x = self.xc + self.radius * np.cos(theta) y = self.yc + self.radius * np.sin(theta) return x, y
[docs] def transformed(self, xfunc=None, yfunc=None): return PolygonalROI(*self.to_polygon()).transformed(xfunc=xfunc, yfunc=yfunc)
[docs] def move_to(self, xdelta, ydelta): self.xc += xdelta self.yc += ydelta
def __gluestate__(self, context): return dict(xc=context.do(self.xc), yc=context.do(self.yc), radius=context.do(self.radius)) @classmethod def __setgluestate__(cls, rec, context): return cls(xc=rec['xc'], yc=rec['yc'], radius=rec['radius'])
[docs]class EllipticalROI(Roi): """ A 2D elliptical region of interest with semimajor/minor axes `radius_[xy]`. Parameters ---------- xc : float, optional `x` coordinate of center. yc : float, optional `y` coordinate of center. radius_x : float, optional Semiaxis along `x` axis. radius_y : float, optional Semiaxis along `y` axis. theta : float, optional Angle of anticlockwise rotation around (`xc`, `yc`) in radian. Notes ----- The `radius_x`, `radius_y` properties refer to the semiaxes along the `x` and `y` axes *before* any rotation is applied. """ def __init__(self, xc=None, yc=None, radius_x=None, radius_y=None, theta=None): super(EllipticalROI, self).__init__() self.xc = xc self.yc = yc self.radius_x = radius_x self.radius_y = radius_y self.theta = 0 if theta is None else theta def __str__(self): if self.defined(): return (f"center=({self.xc:.3f}, {self.yc:.3f}), " f"semiaxes=({self.radius_x:.3f} x {self.radius_y:.3f}), " f"theta={self.theta:.3f} radian") else: return "Undefined Elliptical ROI"
[docs] def contains(self, x, y): if not self.defined(): raise UndefinedROI if not isinstance(x, np.ndarray): x = np.asarray(x) if not isinstance(y, np.ndarray): y = np.asarray(y) if np.isclose(self.theta % np.pi, 0.0, atol=1e-9): return (((x - self.xc) ** 2 / self.radius_x ** 2 + (y - self.yc) ** 2 / self.radius_y ** 2) < 1.) elif np.isclose(self.theta % (np.pi / 2), 0.0, atol=1e-9): return (((x - self.xc) ** 2 / self.radius_y ** 2 + (y - self.yc) ** 2 / self.radius_x ** 2) < 1.) else: # Pre-select points inside the bounding rectangle. In principle this could be # used to speed up the non-rotated cases above as well, but will only pay off # the overhead for datasets much larger than the region (e.g. < ~1 % inside). inside = np.zeros_like(x, dtype=bool) bounds = self.bounds() keep = ((x >= bounds[0][0]) & (x <= bounds[0][1]) & (y >= bounds[1][0]) & (y <= bounds[1][1])) x = x[keep] - self.xc y = y[keep] - self.yc shape = (2,) + x.shape x, y = (rotation_matrix_2d(-self.theta) @ [x.flatten(), y.flatten()]).reshape(shape) inside[keep] = ((x ** 2 / self.radius_x ** 2 + y ** 2 / self.radius_y ** 2) < 1.) return inside
[docs] def reset(self): """Reset the rectangular region""" self.xc = None self.yc = None self.radius_x = 0. self.radius_y = 0.
[docs] def defined(self): return (self.xc is not None and self.yc is not None and self.radius_x is not None and self.radius_y is not None)
[docs] def get_center(self): return self.xc, self.yc
[docs] def to_polygon(self): if not self.defined(): return [], [] theta = np.linspace(0, 2 * np.pi, num=20) x = self.radius_x * np.cos(theta) y = self.radius_y * np.sin(theta) x, y = rotation_matrix_2d(self.theta) @ (x, y) return x + self.xc, y + self.yc
[docs] def bounds(self): """Returns (conservatively estimated) boundary values in `x` and `y`""" if self.theta is None or np.isclose(self.theta % (np.pi), 0.0, atol=1e-9): return [[self.xc - self.radius_x, self.xc + self.radius_x], [self.yc - self.radius_y, self.yc + self.radius_y]] elif np.isclose(self.theta % (np.pi / 2), 0.0, atol=1e-9): return [[self.xc - self.radius_y, self.xc + self.radius_y], [self.yc - self.radius_x, self.yc + self.radius_x]] else: radius = max(self.radius_x, self.radius_y) return [[self.xc - radius, self.xc + radius], [self.yc - radius, self.yc + radius]]
[docs] def transformed(self, xfunc=None, yfunc=None): return PolygonalROI(*self.to_polygon()).transformed(xfunc=xfunc, yfunc=yfunc)
[docs] def move_to(self, xdelta, ydelta): self.xc += xdelta self.yc += ydelta
[docs] def rotate_to(self, theta): self.theta = 0 if theta is None else theta
def __gluestate__(self, context): return dict(xc=context.do(self.xc), yc=context.do(self.yc), radius_x=context.do(self.radius_x), radius_y=context.do(self.radius_y), theta=context.do(self.theta)) @classmethod def __setgluestate__(cls, rec, context): return cls(xc=rec['xc'], yc=rec['yc'], radius_x=rec['radius_x'], radius_y=rec['radius_y'], theta=rec.get('theta', 0))
[docs]class VertexROIBase(Roi): """ Class representing a set of vertices e.g. to define a 2D polygon. Parameters ---------- vx : float or array-like, optional Initial `x` vertices. vy : float or array-like, optional Initial `y` vertices. Notes ----- This class only comprises the collection of vertices, but does not provide a ``contains`` method. """ def __init__(self, vx=None, vy=None): super(VertexROIBase, self).__init__() self.vx = [] if vx is None else list(vx) self.vy = [] if vy is None else list(vy) self.theta = 0
[docs] def transformed(self, xfunc=None, yfunc=None): vx = self.vx if xfunc is None else xfunc(np.asarray(self.vx)) vy = self.vy if yfunc is None else yfunc(np.asarray(self.vy)) return self.__class__(vx=vx, vy=vy)
[docs] def add_point(self, x, y): """ Add another vertex to the ROI. Parameters ---------- x : float The `x` coordinate of the point to add. y : float The `y` coordinate of the point to add. """ self.vx.append(x) self.vy.append(y)
[docs] def reset(self): """Reset the vertex lists and position angle""" self.vx = [] self.vy = [] self.theta = 0
[docs] def replace_last_point(self, x, y): if len(self.vx) > 0: self.vx[-1] = x self.vy[-1] = y
[docs] def remove_point(self, x, y, thresh=None): """ Remove the vertex closest to a reference (`x`, `y`) point. Parameters ---------- x : float The `x` coordinate of the reference point. y : float The `y` coordinate of the reference point. thresh : float, optional Threshold. If set, the vertex closest to (`x`, `y`) will only be removed if the distance is less than `thresh`. """ if len(self.vx) == 0: return # find distance between vertices and input dist = [(x - a) ** 2 + (y - b) ** 2 for a, b in zip(self.vx, self.vy)] inds = range(len(dist)) near = min(inds, key=lambda x: dist[x]) if thresh is not None and dist[near] > (thresh ** 2): return self.vx = [self.vx[i] for i in inds if i != near] self.vy = [self.vy[i] for i in inds if i != near]
[docs] def defined(self): return len(self.vx) > 0
[docs] def to_polygon(self): return self.vx, self.vy
def __gluestate__(self, context): return dict(vx=context.do(np.asarray(self.vx)), vy=context.do(np.asarray(self.vy))) @classmethod def __setgluestate__(cls, rec, context): return cls(vx=context.object(rec['vx']), vy=context.object(rec['vy']))
[docs]class PolygonalROI(VertexROIBase): """ A class to define 2D polygonal regions of interest. Parameters ---------- vx : float or array-like, optional Initial `x` vertices. vy : float or array-like, optional Initial `y` vertices. """ def __str__(self): result = 'Polygonal ROI (' result += ','.join(['(%s, %s)' % (x, y) for x, y in zip(self.vx, self.vy)]) result += ')' return result
[docs] def contains(self, x, y): if not self.defined(): raise UndefinedROI if not isinstance(x, np.ndarray): x = np.asarray(x) if not isinstance(y, np.ndarray): y = np.asarray(y) result = points_inside_poly(x, y, np.asarray(self.vx), np.asarray(self.vy)) return result
# There are several possible definitions of the centre; `mean()` is # easiest to calculate, but not robust against adding vertices.
[docs] def mean(self): """Return arithmetic mean (of vertex positions) of polygon.""" if not self.defined(): raise UndefinedROI # Do not include starting vertex twice! if self.vx[-1] == self.vx[0] and self.vy[:-1] == self.vy[0]: return np.mean(self.vx[:-1]), np.mean(self.vy[:-1]) else: return np.mean(self.vx), np.mean(self.vy)
[docs] def area(self, signed=False): """ Return area of polygon using the shoelace formula. Parameters ---------- signed : bool, optional If `True`, return signed area from the cross product calculation, indicating whether vertices are ordered clockwise (negative) or counter clockwise (positive). """ # Use offsets to improve numerical precision x0, y0 = self.mean() x_ = self.vx - x0 y_ = self.vy - y0 # Shoelace formula; in case where the start vertex is not already duplicated # at the end, final term added manually to avoid an array copy. area_main = np.dot(x_[:-1], y_[1:]) - np.dot(y_[:-1], x_[1:]) if not (self.vx[-1] == self.vx[0] and self.vy[:-1] == self.vy[0]): area_main += x_[-1] * y_[0] - y_[-1] * x_[0] if signed: return 0.5 * area_main else: return 0.5 * np.abs(area_main)
[docs] def centroid(self): """Return centroid (centre of mass) of polygon.""" # See http://paulbourke.net/geometry/polygonmesh/ # https://www.ma.ic.ac.uk/~rn/centroid.pdf # Use vertex position offsets from mean to improve numerical precision; # for a triangle the mean already identifies the centroid. if len(self.vx) == 3: return self.mean() else: x0, y0 = self.mean() if self.vx[-1] == self.vx[0] and self.vy[:-1] == self.vy[0]: x_ = self.vx[:-1] - x0 y_ = self.vy[:-1] - y0 else: x_ = self.vx - x0 y_ = self.vy - y0 indices = np.arange(len(x_)) - 1 xs = x_[indices] + x_ ys = y_[indices] + y_ dxy = x_[indices] * y_ - y_[indices] * x_ scl = 1. / (6 * self.area(signed=True)) return np.dot(xs, dxy) * scl + x0, np.dot(ys, dxy) * scl + y0
[docs] def move_to(self, xdelta, ydelta): self.vx = list(map(lambda x: x + xdelta, self.vx)) self.vy = list(map(lambda y: y + ydelta, self.vy))
[docs] def rotate_to(self, theta, center=None): """ Rotate polygon to position angle `theta` around `center`. Parameters ---------- theta : float Angle of anticlockwise rotation around center in radian. center : pair of float, optional Coordinates of center of rotation. Defaults to :meth:`~glue.core.roi.PolygonalROI.centroid`, for linear "polygons" to :meth:`~glue.core.roi.PolygonalROI.mean`. """ theta = 0 if theta is None else theta # For linear (1D) "polygons" centroid is not defined. if center is None: if self.area() == 0: center = self.mean() else: center = self.centroid() dtheta = theta - self.theta if self.defined() and not np.isclose(dtheta % np.pi, 0.0, atol=1e-9): dx, dy = np.array([self.vx, self.vy]) - np.array(center).reshape(2, 1) self.vx, self.vy = (rotation_matrix_2d(dtheta) @ (dx, dy) + np.array(center).reshape(2, 1)).tolist() self.theta = theta
class Projected3dROI(Roi): """ A region of interest defined in screen coordinates. The screen coordinates are defined by the projection matrix. The projection matrix converts homogeneous coordinates (`x`, `y`, `z`, `w`), where `w` is implicitly 1, to homogeneous screen coordinates (usually the tensor product of the world coordinate vectors and the projection matrix). Parameters ---------- 2d_roi : `~glue.core.roi.Roi`, optional If specified, this ROI will be used in screen coordinate space. projection_matrix : `~numpy.ndarray`, optional Projection matrix defining the mapping from world onto screen coordinates. """ def __init__(self, roi_2d=None, projection_matrix=None): super(Projected3dROI, self).__init__() self.roi_2d = roi_2d self.projection_matrix = np.asarray(projection_matrix) def contains3d(self, x, y, z): if not self.defined(): raise UndefinedROI x = np.asarray(x) y = np.asarray(y) z = np.asarray(z) # Since the projection can significantly increase the memory usage, we # do the following operation in chunks. In future we could likely use # e.g. vaex, dask, or other multi-threaded/fast libraries to speed this # and other ROI code up. mask = np.zeros(x.shape, dtype=bool) for slices in iterate_chunks(x.shape, n_max=1000000): # Work in homogeneous coordinates so we can support perspective # projections as well x_sub, y_sub, z_sub = x[slices], y[slices], z[slices] vertices = np.array([x_sub, y_sub, z_sub, np.ones(x_sub.shape)]) # The following returns homogeneous screen coordinates screen_h = np.tensordot(self.projection_matrix, vertices, axes=(1, 0)) # Convert to screen coordinates, as we don't care about z screen_x, screen_y = screen_h[:2] / screen_h[3] mask[slices] = self.roi_2d.contains(screen_x, screen_y) return mask def __gluestate__(self, context): return dict(roi_2d=context.id(self.roi_2d), projection_matrix=self.projection_matrix.tolist()) @classmethod def __setgluestate__(cls, rec, context): return cls(roi_2d=context.object(rec['roi_2d']), projection_matrix=np.asarray(rec['projection_matrix'])) # TODO: these methods forward directly to roi_2d, not sure if this makes sense for all def contains(self, x, y): return self.roi_2d.contains(x, y) def center(self): return self.roi_2d.center() def move_to(self, x, y): return self.roi_2d.move_to(x, y) def defined(self): return self.roi_2d.defined() def to_polygon(self): return self.roi_2d.to_polygon() def transformed(self, xfunc=None, yfunc=None): return self.roi_2d.transformed(xfunc, yfunc) def rotate_to(self, theta): return self.roi_2d.rotate_to(theta) class Path(VertexROIBase): def __str__(self): result = 'Path (' result += ','.join(['(%s, %s)' % (x, y) for x, y in zip(self.vx, self.vy)]) result += ')' return result
[docs]class AbstractMplRoi(object): """ Base class for objects which use Matplotlib user events to edit/display ROIs. Parameters ---------- axes : :class:`matplotlib.axes.Axes` The Matplotlib axes to draw to. roi : :class:`glue.core.roi.Roi`, optional If specified, this ROI will be used and updated, otherwise a new one will be created. """ _roi_cls = None def __init__(self, axes, roi=None, data_space=True): self._axes = axes self._roi = roi or self._roi_cls() self._previous_roi = None self._mid_selection = False self._scrubbing = False self._background_cache = None self._data_space = data_space def _draw(self): # When drawing the ROI, we first keep a cache of the contents of the # plot then just re-plot the ROI artist on top every time for # performance. However, if the background cache hasn't been set, we need # to do a full draw. if self._background_cache is None or not self._axes.figure.canvas.supports_blit: self._axes.figure.canvas.draw_idle() else: self._axes.figure.canvas.restore_region(self._background_cache) self._axes.draw_artist(self._patch) self._axes.figure.canvas.blit()
[docs] def roi(self): return self._roi.copy()
[docs] def reset(self, include_roi=True): self._mid_selection = False self._scrubbing = False if include_roi: self._roi.reset() self._sync_patch()
[docs] def active(self): return self._mid_selection
[docs] def start_selection(self, event): raise NotImplementedError()
[docs] def update_selection(self, event): raise NotImplementedError()
[docs] def finalize_selection(self, event): raise NotImplementedError()
[docs] def abort_selection(self, event): if self._mid_selection: self._restore_previous_roi() self.reset(include_roi=False)
def _sync_patch(self): raise NotImplementedError() def _store_previous_roi(self): self._previous_roi = self._roi.copy() def _store_background(self): self._background_cache = self._axes.figure.canvas.copy_from_bbox(self._axes.bbox) def _reset_background(self): # The purpose of this method is to provide a way to reset the background # when the figure is changed (e.g. while panning/zooming) while the ROI # is in the middle of being plotted (this is relevant for 'persistent' # ROIs such as path selections or lasso selections). if self._patch is None or not self._patch.get_visible(): return self._background_cache = None self._patch.set_visible(False) self._axes.figure.canvas.draw() self._store_background() self._patch.set_visible(True) self._axes.figure.canvas.draw_idle() def _restore_previous_roi(self): self._roi = self._previous_roi
class MplPickROI(AbstractMplRoi): """ Matplotlib ROI for point selections Parameters ---------- axes : :class:`matplotlib.axes.Axes` The Matplotlib axes to draw to. """ _roi_cls = PointROI def _draw(self): pass def start_selection(self, event): self._roi.x = event.xdata self._roi.y = event.ydata def update_selection(self, event): self._roi.x = event.xdata self._roi.y = event.ydata def finalize_selection(self, event): self._roi.x = event.xdata self._roi.y = event.ydata def _sync_patch(self): pass
[docs]class MplRectangularROI(AbstractMplRoi): """ Matplotlib ROI for rectangular selections Parameters ---------- axes : :class:`matplotlib.axes.Axes` The Matplotlib axes to draw to. """ _roi_cls = RectangularROI def __init__(self, axes, data_space=True): super(MplRectangularROI, self).__init__(axes, data_space=data_space) self._xi = None self._yi = None self.plot_opts = {'edgecolor': PATCH_COLOR, 'facecolor': PATCH_COLOR, 'alpha': 0.3} self._patch = Rectangle((0., 0.), 1., 1., zorder=100) self._patch.set_visible(False) if not self._data_space: self._patch.set_transform(self._axes.transAxes) self._axes.add_patch(self._patch)
[docs] def start_selection(self, event): if event.inaxes != self._axes: return False if self._data_space: xval = event.xdata yval = event.ydata else: axes_trans = self._axes.transAxes.inverted() xval, yval = axes_trans.transform([event.x, event.y]) if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False elif not self._roi.contains(xval, yval): return False self._store_previous_roi() self._store_background() self._xi = xval self._yi = yval if event.key == SCRUBBING_KEY: self._scrubbing = True self._cx, self._cy = self._roi.center() else: self.reset() self._roi.update_limits(self._xi, self._xi, self._yi, self._yi) self._mid_selection = True self._sync_patch() self._draw()
[docs] def update_selection(self, event): if not self._mid_selection or event.inaxes != self._axes: return False if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False if self._data_space: xval = event.xdata yval = event.ydata else: axes_trans = self._axes.transAxes.inverted() xval, yval = axes_trans.transform([event.x, event.y]) if self._scrubbing: self._roi.move_to(self._cx + xval - self._xi, self._cy + yval - self._yi) else: self._roi.update_limits(min(xval, self._xi), min(yval, self._yi), max(xval, self._xi), max(yval, self._yi)) self._sync_patch() self._draw()
[docs] def finalize_selection(self, event): self._scrubbing = False self._mid_selection = False self._patch.set_visible(False) self._draw()
def _sync_patch(self): if self._roi.defined(): corner = self._roi.corner() width = self._roi.width() height = self._roi.height() self._patch.set_xy(corner) self._patch.set_width(width) self._patch.set_height(height) self._patch.set(**self.plot_opts) self._patch.set_visible(True) else: self._patch.set_visible(False) def __str__(self): return "MPL Rectangle: %s" % self._patch
[docs]class MplXRangeROI(AbstractMplRoi): """ Matplotlib ROI for x range selections Parameters ---------- axes : :class:`matplotlib.axes.Axes` The Matplotlib axes to draw to. """ _roi_cls = XRangeROI def __init__(self, axes, data_space=True): super(MplXRangeROI, self).__init__(axes, data_space=data_space) self._xi = None self.plot_opts = {'edgecolor': PATCH_COLOR, 'facecolor': PATCH_COLOR, 'alpha': 0.3} if self._data_space: trans = blended_transform_factory(self._axes.transData, self._axes.transAxes) else: trans = self._axes.transAxes self._patch = Rectangle((0., 0.), 1., 1., transform=trans, zorder=100) self._patch.set_visible(False) self._axes.add_patch(self._patch)
[docs] def start_selection(self, event): if event.inaxes != self._axes: return False if self._data_space: x_val = event.xdata y_val = event.ydata else: transform = self._axes.transAxes.inverted() x_val, y_val = transform.transform([event.x, event.y]) if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False elif not self._roi.contains(x_val, y_val): return False self._store_previous_roi() self._store_background() if event.key == SCRUBBING_KEY: self._scrubbing = True self._dx = x_val - self._roi.center() else: self.reset() self._roi.set_range(x_val, x_val) self._xi = x_val self._mid_selection = True self._sync_patch() self._draw()
[docs] def update_selection(self, event): if not self._mid_selection or event.inaxes != self._axes: return False if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False if self._data_space: xval = event.xdata else: axes_trans = self._axes.transAxes.inverted() xval, _ = axes_trans.transform([event.x, event.y]) if self._scrubbing: self._roi.move_to(xval + self._dx) else: self._roi.set_range(min(xval, self._xi), max(xval, self._xi)) self._sync_patch() self._draw()
[docs] def finalize_selection(self, event): self._scrubbing = False self._mid_selection = False self._patch.set_visible(False) self._draw()
def _sync_patch(self): if self._roi.defined(): rng = self._roi.range() self._patch.set_xy((rng[0], 0)) self._patch.set_width(rng[1] - rng[0]) self._patch.set_height(1) self._patch.set(**self.plot_opts) self._patch.set_visible(True) else: self._patch.set_visible(False)
[docs]class MplYRangeROI(AbstractMplRoi): """ Matplotlib ROI for y range selections Parameters ---------- axes : :class:`matplotlib.axes.Axes` The Matplotlib axes to draw to. """ _roi_cls = YRangeROI def __init__(self, axes, data_space=True): super(MplYRangeROI, self).__init__(axes, data_space=data_space) self._yi = None self.plot_opts = {'edgecolor': PATCH_COLOR, 'facecolor': PATCH_COLOR, 'alpha': 0.3} if self._data_space: trans = blended_transform_factory(self._axes.transAxes, self._axes.transData) else: trans = self._axes.transAxes self._patch = Rectangle((0., 0.), 1., 1., transform=trans, zorder=100) self._patch.set_visible(False) self._axes.add_patch(self._patch)
[docs] def start_selection(self, event): if event.inaxes != self._axes: return False if self._data_space: xval = event.xdata yval = event.ydata else: axes_trans = self._axes.transAxes.inverted() xval, yval = axes_trans.transform([event.x, event.y]) if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False elif not self._roi.contains(xval, yval): return False self._store_previous_roi() self._store_background() if event.key == SCRUBBING_KEY: self._scrubbing = True self._dy = yval - self._roi.center() else: self.reset() self._roi.set_range(yval, yval) self._yi = yval self._mid_selection = True self._sync_patch() self._draw()
[docs] def update_selection(self, event): if not self._mid_selection or event.inaxes != self._axes: return False if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False if self._data_space: yval = event.ydata else: axes_trans = self._axes.transAxes.inverted() _, yval = axes_trans.transform([event.x, event.y]) if self._scrubbing: self._roi.move_to(yval + self._dy) else: self._roi.set_range(min(yval, self._yi), max(yval, self._yi)) self._sync_patch() self._draw()
[docs] def finalize_selection(self, event): self._scrubbing = False self._mid_selection = False self._patch.set_visible(False) self._draw()
def _sync_patch(self): if self._roi.defined(): rng = self._roi.range() self._patch.set_xy((0, rng[0])) self._patch.set_height(rng[1] - rng[0]) self._patch.set_width(1) self._patch.set(**self.plot_opts) self._patch.set_visible(True) else: self._patch.set_visible(False)
[docs]class MplCircularROI(AbstractMplRoi): """ Matplotlib ROI for circular selections Since circles on the screen may not be circles in the data (due, e.g., to logarithmic scalings on the axes), the ultimate ROI that is created is a polygonal ROI Parameters ---------- axes : :class:`matplotlib.axes.Axes` The Matplotlib axes to draw to. """ _roi_cls = CircularROI def __init__(self, axes, data_space=True): super(MplCircularROI, self).__init__(axes, data_space=data_space) self.plot_opts = {'edgecolor': PATCH_COLOR, 'facecolor': PATCH_COLOR, 'alpha': 0.3} self._xi = None self._yi = None self._patch = Ellipse((0., 0.), transform=IdentityTransform(), width=0., height=0., zorder=100) self._patch.set_visible(False) self._axes.add_patch(self._patch) def _sync_patch(self): if self._roi.defined(): xy = self._roi.get_center() r = self._roi.get_radius() self._patch.center = xy self._patch.width = 2. * r self._patch.height = 2. * r self._patch.set(**self.plot_opts) self._patch.set_visible(True) else: self._patch.set_visible(False)
[docs] def start_selection(self, event): if event.inaxes != self._axes: return False xy = data_to_pixel(self._axes, [event.xdata], [event.ydata]) xi = xy[0, 0] yi = xy[0, 1] if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False elif not self._roi.contains(xi, yi): return False self._store_previous_roi() self._store_background() if event.key == SCRUBBING_KEY: self._scrubbing = True (xc, yc) = self._roi.get_center() self._dx = xc - xi self._dy = yc - yi else: self.reset() self._roi.set_center(xi, yi) self._roi.set_radius(0.) self._xi = xi self._yi = yi self._mid_selection = True self._sync_patch() self._draw()
[docs] def update_selection(self, event): if not self._mid_selection or event.inaxes != self._axes: return False xy = data_to_pixel(self._axes, [event.xdata], [event.ydata]) xi = xy[0, 0] yi = xy[0, 1] if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False if self._scrubbing: self._roi.set_center(xi + self._dx, yi + self._dy) else: dx = xy[0, 0] - self._xi dy = xy[0, 1] - self._yi self._roi.set_radius(np.hypot(dx, dy)) self._sync_patch() self._draw()
[docs] def roi(self): if not self._roi.defined(): return PolygonalROI() # Get the circular ROI parameters in pixel units xy_center = self._roi.get_center() rad = self._roi.get_radius() # At this point, if one of the axes is not linear, we convert to a polygon if (self._axes.get_xscale() != 'linear' or self._axes.get_yscale() != 'linear') and self._data_space: theta = np.linspace(0, 2 * np.pi, num=200) x = xy_center[0] + rad * np.cos(theta) y = xy_center[1] + rad * np.sin(theta) xy_data = pixel_to_data(self._axes, x, y) vx = xy_data[:, 0].ravel().tolist() vy = xy_data[:, 1].ravel().tolist() result = PolygonalROI(vx, vy) else: # We should now check if the radius in data coordinates is the same # along x and y, as if so then we can return a circle, otherwise we # should return an ellipse. x = xy_center[0] + np.array([0, 0, rad]) y = xy_center[1] + np.array([0, rad, rad]) xy_data = pixel_to_data(self._axes, x, y) if self._data_space else pixel_to_axes(self._axes, x, y) rx = xy_data[2, 0] - xy_data[0, 0] ry = xy_data[1, 1] - xy_data[0, 1] xc, yc = xy_data[0, :] if np.allclose(rx, ry): return CircularROI(xc=xc, yc=yc, radius=rx) else: return EllipticalROI(xc=xc, yc=yc, radius_x=rx, radius_y=ry) return result
[docs] def finalize_selection(self, event): self._scrubbing = False self._mid_selection = False self._patch.set_visible(False) self._draw()
[docs]class MplPolygonalROI(AbstractMplRoi): """ Matplotlib ROI for polygon selections Parameters ---------- axes : :class:`matplotlib.axes.Axes` The Matplotlib axes to draw to. roi : :class:`glue.core.roi.Roi`, optional If specified, this ROI will be used and updated, otherwise a new one will be created. """ _roi_cls = PolygonalROI def __init__(self, axes, roi=None, data_space=True): super(MplPolygonalROI, self).__init__(axes, roi=roi, data_space=data_space) self.plot_opts = {'edgecolor': PATCH_COLOR, 'facecolor': PATCH_COLOR, 'alpha': 0.3} self._patch = Polygon(np.array(list(zip([0, 1], [0, 1]))), zorder=100) self._patch.set_visible(False) if not self._data_space: self._patch.set_transform(self._axes.transAxes) self._axes.add_patch(self._patch) def _sync_patch(self): if self._roi.defined(): x, y = self._roi.to_polygon() self._patch.set_xy(list(zip(x + [x[0]], y + [y[0]]))) self._patch.set_visible(True) self._patch.set(**self.plot_opts) else: self._patch.set_visible(False)
[docs] def start_selection(self, event, scrubbing=False): if event.inaxes != self._axes: return False if self._data_space: xval = event.xdata yval = event.ydata else: axes_trans = self._axes.transAxes.inverted() xval, yval = axes_trans.transform([event.x, event.y]) if scrubbing or event.key == SCRUBBING_KEY: if not self._roi.defined(): return False elif not self._roi.contains(xval, yval): return False self._store_previous_roi() self._store_background() if scrubbing or event.key == SCRUBBING_KEY: self._scrubbing = True self._cx = xval self._cy = yval else: self.reset() self._roi.add_point(xval, yval) self._mid_selection = True self._sync_patch() self._draw()
[docs] def update_selection(self, event): if not self._mid_selection or event.inaxes != self._axes: return False if event.key == SCRUBBING_KEY: if not self._roi.defined(): return False if self._data_space: xval = event.xdata yval = event.ydata else: axes_trans = self._axes.transAxes.inverted() xval, yval = axes_trans.transform([event.x, event.y]) if self._scrubbing: self._roi.move_to(xval - self._cx, yval - self._cy) self._cx = xval self._cy = yval else: self._roi.add_point(xval, yval) self._sync_patch() self._draw()
[docs] def finalize_selection(self, event): self._scrubbing = False self._mid_selection = False self._patch.set_visible(False) self._draw()
class MplPathROI(MplPolygonalROI): """ Matplotlib ROI for path selections Parameters ---------- axes : :class:`matplotlib.axes.Axes` The Matplotlib axes to draw to. """ _roi_cls = Path def __init__(self, axes, roi=None): super(MplPolygonalROI, self).__init__(axes) self.plot_opts = {'edgecolor': PATCH_COLOR, 'facecolor': PATCH_COLOR, 'alpha': 0.3} self._patch = None def start_selection(self, event): if self._patch is not None: self._patch.remove() self._patch = None self._background_cache = None self._axes.figure.canvas.draw() super(MplPathROI, self).start_selection(event) def _sync_patch(self): if self._patch is not None: self._patch.remove() self._patch = None if self._roi.defined(): x, y = self._roi.to_polygon() p = MplPath(np.column_stack((x, y))) self._patch = PathPatch(p, transform=self._axes.transData) self._patch.set_visible(True) self._patch.set(**self.plot_opts) self._axes.add_artist(self._patch) def finalize_selection(self, event): self._mid_selection = False if self._patch is not None: self._patch.remove() self._patch = None self._draw()
[docs]class CategoricalROI(Roi): """ A ROI abstraction to represent selections of categorical data. """ def __init__(self, categories=None): if categories is None: self.categories = None else: self.update_categories(categories)
[docs] def to_polygon(self): """ Just not possible. """ raise NotImplementedError
def _categorical_helper(self, indata): """ A helper function to do the rigamaroll of getting categorical data. Parameters ---------- indata : object Any type of input data Returns ------- The best guess at the categorical data associated with indata. """ try: if isinstance(indata, CategoricalComponent): return indata.data else: return indata[:] except AttributeError: return np.asarray(indata)
[docs] def contains(self, x, y): """ Test whether a set categorical elements fall within the region of interest. Parameters ---------- x : array-like An array-like object of categories (includes `CategoricalComponents`). y : object or None Unused but required for compatibility Returns ------- A list of True/False values, for whether each `x` value falls within the ROI """ if self.categories is None or len(self.categories) == 0: return np.zeros(x.shape, dtype=bool) else: check = self._categorical_helper(x) index = np.minimum(np.searchsorted(self.categories, check), len(self.categories) - 1) return self.categories[index] == check
[docs] def update_categories(self, categories): self.categories = np.unique(self._categorical_helper(categories))
[docs] def defined(self): return self.categories is not None
[docs] def reset(self): self.categories = None
[docs] @staticmethod def from_range(categories, lo, hi): """ Utility function to help construct the ROI from a range. Parameters ---------- categories : object Anything understood by ``._categorical_helper`` ... array, list or component. lo : int or float Lower bound of the range (rounded up to next integer) hi : int or float Upper bound of the range (rounded up to next integer) Returns ------- `CategoricalROI` object """ # Convert lo and hi to integers. Note that if lo or hi are negative, # which can happen if the user zoomed out, we need to reset the to zero # otherwise they will have strange effects when slicing the categories. # Note that we used ceil for lo, because if lo is 0.9 then we should # only select 1 and above. lo = np.intp(np.ceil(lo) if lo > 0 else 0) hi = np.intp(np.ceil(hi) if hi > 0 else 0) roi = CategoricalROI() roi.update_categories(categories[lo:hi]) return roi
def __gluestate__(self, context): return dict(categories=self.categories.tolist()) @classmethod def __setgluestate__(cls, rec, context): return cls(categories=rec['categories'])